May 27, 2023 RL + Model-based Control: Using On-demand Optimal Control to Learn Versatile Legged Locomotion